Keywords Are Dead, Do THIS Instead.

Author: Caleb Ulku | Published: 2026-08-04 | Source: transcript (28:51, 25.1K views at ingest)

Summary

A reaction video: Caleb Ulku plays Neil Patel on “keywords are dead / Google micro-moments,” agrees with most of it, then argues Neil skips the local-money moment. Neil recaps Google’s 2015 micro-moments (know / go / do / buy) and covers three; Caleb says I want to go — plumber, roofer, med spa — gets zero seconds, “not even a chapter.” Both agree the keyword era is over: old SEO was exact-match matching, not understanding the user. Neil cites majority zero-click already pre-AIO (57% → 59%). Caleb’s analogue is a client who ranked #1 for “mortgage calculator” until Google put a calculator on the SERP; AI Overviews do that to informational queries.

His claims sit on Entity-Based SEO and GEO / AEO (Getting Recommended by AI), not classical topical authority. If a local agency leads with keywords, hire someone else. People describe whole situations to ChatGPT; models match attributes on websites and reviews. Ranking #1 for “plumber near me” has “no impact” on a “garbage disposal making a funny noise” recommendation. He has found no AI-visibility tracker worth paying for: four “same intent” prompts produced zero overlapping recs; the same prompt twice often differs; chat memory adds untracked context. He uses GA AI-referral traffic as a noisy proxy (it misses “AI names the brand → user Googles it”).

The content he wants is trust content — pricing, problems, not-a-fit, repair-vs-replace — localized (tankless caveats differ Minneapolis vs. Miami). Humans spent <30s on local sites, so this never mattered much for Google; AIs read it and treat it as local. He demos a new “trust” mode in the Core 30 agent (~80 prompts: research, outline, cross-model fact-check, write, Schema Markup (Structured Data), humanize, QA). He agrees with Neil’s on-page tactics (lead with the answer; question-as-heading; spoken-language FAQ) and maps “lead with the answer” to Google’s goal completion. History-first homepages fail users and AI.

Off-site: Reddit is still cited but “a lot less than a year ago” because SEOs spammed it (preview of They Lied About Reddit (Google Proved It)). Gemini Ask Maps cares about review text and about Yelp / Angie’s List, not just the Google Business Profile. He rotates platforms and asks for outcome-specific reviews (inside Google’s no-incentive rules). GBP is the entity heart; ChatGPT uses Bing for Business, so copy GBP there. He is “surprised” Neil still pushes roundups — current models barely use them. For local, on-page trust content beats national mentions because competitors aren’t doing it.

Prediction: Ask Maps (March) is a harbinger of Gemini entering the three-pack. Gemini cares much less about proximity (Local Ranking Factors (Relevance, Proximity, Prominence/Authority)) and more about attributes; businesses ranking on proximity alone will get hurt. Neil’s “AI-search leads 3.1% → 7.4%” — Caleb agrees on direction, still a fraction of three-pack volume. Dual plugs: Neil owns Answer the Public; Caleb sells Core 30.

Key Claims

Keywords / keyword research are “dead” for how Google and AI systems work; do not hire a local agency that leads with keywords. (Caleb + Neil, creator-opinion)
Google’s 2015 micro-moments (know / go / do / buy) still apply; Neil skips I want to go, which is the local-money query.
Zero-click was already majority pre-AIO (Neil: 57% → 59%); AIO sponges informational CTR the way a SERP calculator killed “mortgage calculator.”
AI-visibility trackers are not actionable: prompt specificity, non-determinism, and chat memory make “same intent” tracking meaningless. Use GA AI-referrals as a noisy proxy.
Attribute matching: models read site copy and reviews for the specific situation (garbage-disposal noise, tankless doubts), not the head term “plumber near me.”
Trust content (pricing / problems / not-a-fit / repair-vs-replace), localized, is how local businesses get cited — newly productized in Core 30’s ~80-prompt pipeline.
Lead with the answer (Google goal completion); question-as-heading; FAQ in user language. History-first homepages fail users and AI.
Gemini Ask Maps uses review text + non-GBP sources (Yelp, Angie’s); rotate platforms; ask for outcome-specific reviews. Copy GBP → Bing for Business (ChatGPT’s local graph).
Roundups are largely ignored by current models (Caleb vs. Neil). For local, on-page > national mentions.
Prediction: Gemini will enter the map pack; proximity down, attributes up — a decade-scale change.

Notable quotes

“Keywords are dead. Keyword research is dead. That’s just not how Google works anymore. It’s not how the AI systems work anymore either.”

Why citable: The title claim at full strength — useful as a positioning line, and the one to keep tagged creator-opinion rather than established fact.

“The fact that you’re ranking number one for plumber near me has no impact on where you’re ranking when somebody types into ChatGPT that their garbage disposal is making a funny noise.”

Why citable: Sharpest illustration of attribute matching vs. head-term SEO; operationalizes Entity-Based SEO for GEO / AEO (Getting Recommended by AI).

“Gemini does not care nearly as much about proximity… instead of proximity, it cares about attributes.”

Why citable: The map-pack prediction in one line — a testable, dated (post–Ask Maps, March) hypothesis against the established proximity/relevance/authority triad.

Connections

Entities mentioned: Caleb Ulku, Google, ChatGPT, Claude, Gemini, Reddit, Google Business Profile, Core30 AI Tool · Neil Patel / Answer the Public (NEW), Yelp, Angie’s List, Bing for Business, Seer Interactive (minor) Concepts referenced: Entity-Based SEO, GEO / AEO (Getting Recommended by AI), Local SEO, Local Ranking Factors (Relevance, Proximity, Prominence/Authority), Core 30, 8-Pass AI Writing Pipeline (the multi-step write/fact-check/humanize pipeline), Review & Reputation Management, Schema Markup (Structured Data), GBP–Website Alignment (Consistency Signals) (GBP as entity heart; Bing copy)

Contradictions / Tensions

“Keywords are dead” vs. his own Core 30. The canonical tutorial still targets category+city and service+city as the page keywords. This video is a shift in what you optimize for (attributes/moments inside those pages), not a retraction of the page inventory. Don’t let the slogan erase the system.
vs. Local Ranking Factors (Relevance, Proximity, Prominence/Authority) proximity. He predicts Gemini-in-the-pack will downgrade proximity. That is a forward-looking claim against the established triad; keep it emerging until it happens.
vs. Neil on roundups / off-site 75%. Caleb says current models ignore abused roundups and that on-page trust content is the local lever; Neil’s audience is large SaaS, where off-site mentions matter more. Named audience split, not a silent average.
vs. TRAIN Any AI To Recommend You (This Study Proves How) Reddit-as-primary-source: already softening here (“a lot less than a year ago”); full correction in They Lied About Reddit (Google Proved It) a week later.
vs. “never write blog posts” (The BEST SEO Tutorial for Businesses 2026 (Full Guide) (saved as 'The BEST Local SEO Tutorial for Beginners 2026')): trust articles (problems, pricing, not-a-fit) are exactly the blog-shaped content he previously told local owners to run from. Soften the old slogan; this is a real shift, productized as Core 30 “trust” type.
Dual commercial payload: Neil → Answer the Public; Caleb → Core 30 agent. Tracker-skepticism is convenient if he doesn’t sell a tracker; still a fair methodological point.

Notes

Not topical-authority. The “keywords are dead” thesis here is moment / attribute / entity matching (and localized trust content so AIs can cite you). Classical topical authority (own the cluster, cover every subtopic to rank the head term) is not what he argues. Do not mint a topical-authority page from this video.

No grumpy-seo-guy. Ahrefs not mentioned. Watermarking not mentioned (see Claude's Watermarks Just Broke SEO, 10 days later).

Neil’s cited figures (Seer CTR, 57/59% zero-click, 31.4% AI-citation-at-rank-1, 90% of cited pages rank 21+, 3.1%→7.4% AI leads) are Neil’s numbers, relayed. Transcript garbles at least one (1.76% “down to 61%” vs. “more than a 60% drop”). Backfill from Neil’s original before treating them as data.